摘要
针对多指标面板数据的样品分类问题,从特征提取角度提出一个多指标面板数据的聚类分析方法。该方法将时间序列的局部变化特性与整体距离关系结合起来,将局部变化的信息融入相似测度的计算中,提出一种自适应滑动窗口分段方法,实现时间序列局部变化的特征提取,在重新定义综合距离的基础上,提出一种聚类方法。通过实证分析,表明新方法能够解决多指标面板数据聚类的问题,分类效果较好。
To address the problem of sample classification of multivariate panel data, this paper proposes a clustering approach based on shape. It can comprehensively consider the panel data＇s local changes characteristics in time series dimension with global distance. It proposes an adaptive sliding windows section method to implement the shape extraction. Based on reconstructing the synthesized distance, it presents a panel data clustering analysis method. The empirical analysis shows that this method can solve the problem of panel data clustering, and the clustering results shows good applicability.
出处
《统计与信息论坛》
CSSCI
2011年第10期28-33,共6页
Journal of Statistics and Information
基金
国家社会科学基金重点项目<我国制造企业创新型成本领先战略研究>(11AGL001)
